Committee |
Date Time |
Place |
Paper Title / Authors |
Abstract |
Paper # |

IBISML |
2022-03-08 10:55 |
Online |
Online |
Real log canonical threshold of reduced rank regression when inputs are on a low dimensional hyperplane Joe Hirose, Sumio Watanabe (Tokyo Tech) IBISML2021-32 |
A reduced rank regression is a statistical model which estimates a linear regression function from in- puts to outputs w... [more] |
IBISML2021-32 pp.15-18 |

NC, MBE (Joint) |
2020-03-05 13:00 |
Tokyo |
University of Electro Communications (Cancelled but technical report was issued) |
Bayesian learning curve for the case when the optimal distribution is not unique Shuya Nagayasu, Sumio Watanabe (Tokyo Tech) NC2019-94 |
Bayesian inference is a widely used statistical method. Asymptotic behaviors of generalization loss and free energy in B... [more] |
NC2019-94 pp.107-112 |

IBISML |
2020-01-09 13:00 |
Tokyo |
ISM |
Asymptotic Behavior of Bayesian Generalization Error in Multinomial Mixtures Takumi Watanabe, Sumio Watanabe (Tokyo Tech) IBISML2019-18 |
Multinomial mixtures are widely used in the information engineering field. However, it is not subject to the conventiona... [more] |
IBISML2019-18 pp.1-8 |

IBISML |
2020-01-09 13:25 |
Tokyo |
ISM |
Real Log Canonical Threshold of Three Layered Neural Network with Swish Activation Function Raiki Tanaka, Sumio Watanabe (Tokyo Tech) IBISML2019-19 |
In neural network learning, it is known that selection of activation function effects generalization performance. Althou... [more] |
IBISML2019-19 pp.9-15 |

MBE, NC (Joint) |
2018-03-14 10:25 |
Tokyo |
Kikai-Shinko-Kaikan Bldg. |
Experimental Analysis of Real Log Canonical Threshold in Stochastic Matrix Factorization using Hamiltonian Monte Carlo Method Naoki Hayashi, Sumio Watanabe (Tokyo Tech) NC2017-89 |
For the real log canonical threshold (RLCT) that gives the Bayesian generalization error of stochastic matrix factorizat... [more] |
NC2017-89 pp.127-131 |

IBISML |
2018-03-05 13:00 |
Fukuoka |
Nishijin Plaza, Kyushu University |
Real Log Canonical Threshold and Bayesian Generalization Error of Mixture of Poisson Distributions Kenichiro Sato, Sumio Watanabe (Tokyo Inst. of Tech.) IBISML2017-90 |
[more] |
IBISML2017-90 pp.1-6 |

IBISML |
2017-11-09 13:00 |
Tokyo |
Univ. of Tokyo |
[Poster Presentation]
Real Log Canonical Threshold of Stochastic Matrix Factorization and its Application to Bayesian Learning Naoki Hayashi, Sumio Watanabe (TokyoTech) IBISML2017-38 |
In stochastic matrix factorization (SMF), we deal with problems that we predict an observed stochastic matrix as a produ... [more] |
IBISML2017-38 pp.23-30 |

MBE, NC (Joint) |
2017-03-13 10:00 |
Tokyo |
Kikai-Shinko-Kaikan Bldg. |
Experimental Analysis of Real Log Canonical Threshold in Non-negative Matrix Factorization Naoki Hayashi, Sumio Watanabe (Tokyo Tech) NC2016-78 |
For the real log canonical threshold ( RLCT ) that gives the Bayesian generalization error of non-negative matrix factor... [more] |
NC2016-78 pp.85-90 |

IBISML |
2016-11-16 15:00 |
Kyoto |
Kyoto Univ. |
Estimation of Vehicular Headway-Velocity Characteristics in Mixture of Piecewise Linear Model using Variational Bayes Method Fumito Nakamura, Sumio Watanabe (Tokyo Tech) IBISML2016-65 |
[more] |
IBISML2016-65 pp.137-142 |

IBISML |
2016-11-16 15:00 |
Kyoto |
Kyoto Univ. |
[Poster Presentation]
Optimization Method of Deep Ensemble Learning using Hierarchical Clustering Natsuki Koda, Sumio Watanabe (Tokyo Tech) IBISML2016-70 |
The method which is used for prediction by combining many different learning machines generated by using same training d... [more] |
IBISML2016-70 pp.171-176 |

IBISML |
2016-11-17 14:00 |
Kyoto |
Kyoto Univ. |
[Poster Presentation]
A real log canonical threshold of nonnegative matrix factorization and its application to Bayesian learning Naoki Hayashi, Sumio Watanabe (Tokyo Tech) IBISML2016-76 |
In nonnegative matrix factorization(NMF)，we deal with problems that we predict a data matrix as a product of two
nonne... [more] |
IBISML2016-76 pp.215-220 |

MBE, NC (Joint) |
2016-03-23 13:10 |
Tokyo |
Tamagawa University |
Evaluation Method of Free Energy Calculation by Replica Monte Carlo Method using Nongaussian and Solvable Models Shoji Sugai, Sumio Watanabe (Tokyo Tech) NC2015-83 |
[more] |
NC2015-83 pp.77-82 |

IBISML |
2016-03-17 15:45 |
Tokyo |
Institute of Statistical Mathematics |
Learning and Generalization in Neural Networks using Hamiltonian Monte Carlo Method Fumito Nakamura, Sumio Watanabe (Tokyo Tech) IBISML2015-97 |
[more] |
IBISML2015-97 pp.25-29 |

IBISML |
2015-11-26 15:00 |
Ibaraki |
Epochal Tsukuba |
[Poster Presentation]
Classification of Training Results in Nonlinear Multi-Layer Principal Component Analysis using Sparse Representation Natsuki Koda, Sumio Watanabe (Tokyo Tech) IBISML2015-55 |
The bottleneck neural network or Nonlinear Multi-Layer Principal Component Analysis(NMPCA) is used to extract the low di... [more] |
IBISML2015-55 pp.19-24 |

NC, MBE |
2015-03-17 13:50 |
Tokyo |
Tamagawa University |
Optimization of LASSO Learning using WAIC and Its Application to City Data Analysis Dai Miyazaki, Sumio Watanabe (Tokyo Tech) MBE2014-175 NC2014-126 |
LASSO(Least Absolute Shrinkage and Selection Operator) is a method adding a penalty term consisting of absolute values o... [more] |
MBE2014-175 NC2014-126 pp.331-336 |

IBISML |
2014-11-18 15:00 |
Aichi |
Nagoya Univ. |
[Poster Presentation]
Optimization Method of LASSO Hyperparameter using WAIC Dai Miyazaki, Sumio Watanabe (Tokyo Tech) IBISML2014-63 |
LASSO (Least Absolute Shrinkage and Selection Operator) was proposed as a regression method using a penalty term made of... [more] |
IBISML2014-63 pp.213-218 |

SP, IPSJ-SLP (Joint) |
2014-07-25 13:20 |
Iwate |
Hotel Hanamaki |
[Invited Talk]
Evaluation Criteria of Statistical Learning when Gaussian Approximation can not be Applied to Likelihood Function Sumio Watanabe (Tokyo Inst. of Tech.) SP2014-68 |
Conventional statistical asymptotic theory was established based on the assumption that the likelihood function can be a... [more] |
SP2014-68 pp.31-36 |

NC, MBE (Joint) |
2014-03-18 13:40 |
Tokyo |
Tamagawa University |
Computational validation of the information criterion WBIC by the exchange Monte Carlo method Satoru Tokuda, Kenji Nagata (Univ. of Tokyo), Sumio Watanabe (Tokyo Inst. of Tech.), Masato Okada (Univ. of Tokyo/RIKEN) NC2013-109 |
In the models with hierarchy like artificial neural networks and mixture models, asymptotic normality, which AIC and BIC... [more] |
NC2013-109 pp.121-126 |

NC, MBE (Joint) |
2013-12-21 13:30 |
Gifu |
Gifu University |
Difference of Enough Nmbers for General and Regular Asymptotic Theories in Statistical Learning Sumio Watanabe (Tokyo Tech) NC2013-61 |
There are two asymptotic theories in statistical learning. One is the regular theory which assumes that the likelihood f... [more] |
NC2013-61 pp.47-52 |

IBISML |
2013-11-12 15:45 |
Tokyo |
Tokyo Institute of Technology, Kuramae-Kaikan |
[Poster Presentation]
Model Selection of Layered Neural Networks using WBIC based on Steepest Descent and MCMC Method Yusuke Tamai, Sumio Watanabe (Tokyo Inst. of Tech.) IBISML2013-36 |
Many learning machines such as neural networks, normal mixtures, and hidden Markov Models contain hierarchical layers, h... [more] |
IBISML2013-36 pp.1-6 |